AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[EsaT]

Tested does not mean validated. The AMD reps quoted in the Toms article have no idea what they're talking about.
Part of the qualification is proof by design, and by definition no PCIe 3.0 mobo is going to meet the 4.0 design requirements.

This stuff is way way waayyy more complex than average joe realises.

Manufacturers (should) know how well they designed and built their boards.
If it was designed for just passing minimums, there's certainly no hope, but if design was for high quality...

And if not possible at all, why manufacturers even had PCIe v4 in those available for public Ryzen 3000 support beta BIOSes?
They would have kept it in internal testing if chance for working was small.
Common sense should tell that.
Or I guess neither mobo makers have any idea of what they're doing...

 

[DragonQ]

I think single chiplet will beat two chiplets all the time in games.

We won't see single chiplet designs until Renoir though, right?

 

[beany_bot]

We won't see single chiplet designs until Renoir though, right?

The 4, 6 and 8 core Zen2 CPU's are single chiplet.

 

[RavenXXX2]

I'm going for the 3600 to slot in to a B450, hope they can do at least 4.4Ghz. Would love to see it beating out the 8700k at nearly half the price.

 

[humbug]

5 Days

 

[Hagar]

Most probably your 1st gen holding back your ram. My Ripjaw V 3200 CL16 won't run DOCP on 1st gen but it can on my 2nd gen using same motherboard. It can even oc to 3466 CL16.

My Ripjaw V 3200 CL14 will run DOCP on 1st gen. It will also run stilts fast timings and will OC to 3333 CL14

 

[Zeeflyboy]

I think single chiplet will beat two chiplets all the time in games.

Maybe. Some games will use more than 8 cores, so even if there is a latency penalty the 8 core is unlikely to better the 12 in such games (though they are definitely very few and far between at the moment, the only one I can think of that I personally play is star citizen... but we may see more over the next few years?).

There is also a possibility that the 12 core could overclock higher than the 8 core, given each chiplet will only have 6 cores running vs the 8 in the single one - less heat in a given area.

Just too many unknowns to call it either way at the moment in my opinion.

In terms of gaming performance to cost ratio, the 8 core will undoubtedly be substantially better than the 12 core, especially given that I'm sure both will cause the GPU to be the bottleneck for typical gamers. The latter may be better performance value if your rig both games and does other work that can leverage all those cores (rendering etc).

 

[DragonQ]

The 4, 6 and 8 core Zen2 CPU's are single chiplet.

Link please?

 

[EsaT]

That will largely depends on how AMD have setup the L3 cache, if each chiplet can share it's cache with the other that will give it a huge advantage over a single chiplet design.

That's the big question mark.
Can L3 of other chiplet be utilized in some way and what's its latency compared to RAM.

Also with chance to use only better cores of the dies, six cores per die could clock better/at lower volts.
So there are multiple variables.

 

[Orange Nexus]

My Ripjaw V 3200 CL14 will run DOCP on 1st gen. It will also run stilts fast timings and will OC to 3333 CL14

That's the manuffacturer's spec? If it is then its b-die. Mine is 3200 Cl16. M-die. And it does 3200 Cl16 DOCP but crashes in games. My Flares 3200 Cl14 can run DOCP and oc just as good on my 1st gen.

 

[4K8KW10]

Did you even look at which system this heatshield was on? It was an Intel Z270 board, not an AMD board, and this is an AMD thread.

Also, i have the MSI Carbon X470 and the heatshield i have on this X470 looks totally different.

Stop speading twisted crap in here.

My M.2 never sees over 53c under its heatshield and a Vega 56 GPU.. just so you know

Your observations don't change the fact that msi's boards are worse than Asus's, Gigabyte's and ASrock's.

 

[beany_bot]

Link please?

em, it's common knowledge. Here is Lisa showing off the 8 core Zen 2 part.

 

[MyBrainz]

the fact that msi's boards are worse than Asus's, Gigabyte's and ASrock's.

You obviously just dont like MSI, i have never had a problem with them, but to say the above is just ridiculous. I thought we were talking about the heatshield for the M.2.. and you have jumped right into the whole board is worse than the 3 you quoted above.

 

[Potatowithearsontheside]

I don't buy the logic of a 12c dual chiplet overclocking higher than an 8c single chiplet due to only 6cbeing active per chiplet. If that were the case then it should hold true for 6c Zen 2 and 4/6c Zen/Zen+. There's no evidence that they do, so I think it a nonsense to suggest that the 12c would clocl higher for that stated reason.
There may be other reasons for 12c to clock higher, read binning, but not the reason listed above.

 

[Journey]

I don't buy the logic of a 12c dual chiplet overclocking higher than an 8c single chiplet due to only 6cbeing active per chiplet. If that were the case then it should hold true for 6c Zen 2 and 4/6c Zen/Zen+. There's no evidence that they do, so I think it a nonsense to suggest that the 12c would clocl higher for that stated reason.
There may be other reasons for 12c to clock higher, read binning, but not the reason listed above.

Spreading the heat out over a larger surface area has a big effect and since two cores on each chiplet would not be active there is a possibility that the saved power and better heat dissipation would allow better clocks.

 

[DragonQ]

Spreading the heat out over a larger surface area has a big effect and since two cores on each chiplet would not be active there is a possibility that the saved power and better heat dissipation would allow better clocks.

But as Mr Potato said, that wasn't the case with Ryzen 1 or 2. The problem is Ryzen overclocking has so far been voltage limited rather than temperature limited. If it was temperature limited, one could argue that fewer cores = less heat = more headroom, but the voltage wall is the real problem. Zen 2 might change that but we don't know yet.

 

[Zeeflyboy]

But it isn’t necessary to be so binary... it doesn’t have to be either a well binned chiplet or 6 cores per chiplet.

If you have two otherwise equally good chiplets, the one with only 6 cores active will have a lower heat per unit area than the 8 core, that is literally all I am suggesting. So given that the 3800x and 3900X are the current top tier of their core counts, it stands to reason they are both possibly using the best binned cores for Ryzen3000... all else being equal, the 6 core chiplet of the 12 core will have lower thermal density which might allow higher over locking headroom. That is literally the only suggestion I made - it’s a possibility, nothing more.

The lower tier 6 cores are perhaps more likely to be taken from less well performing bins, it’s perfectly conceivable that the 12 core will be utilising chiplets from a higher bin...

 

[Potatowithearsontheside]

Theory and practice commonly diverge, often because you hit another bottleneck before that theory can become reality.

 

[awaybreaktoday]

em, it's common knowledge. Here is Lisa showing off the 8 core Zen 2 part.

Each die is two chiplets.... Except the I/O die.

 

[Zeeflyboy]

Theory and practice commonly diverge, often because you hit another bottleneck before that theory can become reality.

Quite true but I think to rule it out as decisively as you did is incorrect. It’s a perfectly reasonable possibility.

Given that the 3900X actually has a 100mhz higher boost than the 3800X according to AMD’s specs I think there is a reasonable chance that if you put enough cooling on there to remove the heat from the extra core count you may be able to squeeze a bit more clock out of it than the 3800X before running into the bottleneck of the thermal density.

I’m going to bet that given sufficient cooling and VRMs the typical 3900X will Overclock a little higher than the typical 3800X. I could well be wrong, time will tell.